HPUX client index database format problem


Greetings:

I've got a working Amanda backup system using amanda-2.4.4p1. Clients
and server are all RedHat Enterprise Linux 3 (RHEL3) machines, and I'm 
using the amanda-2.4.4p1-0.3E RPMs that came with RHEL3.

I've recently added HPUX-11.0 clients to the system, compiling
amanda-2.4.4p3 from source for the HPUX clients.  The HPUX clients
appear to be working, with one exception; when running amrecover on
the amanda server trying to recover an HPUX client an error 
"No index records for disk for specified date" occurs:

        amrecover> sethost seagull
        200 Dump host set to seagull.
        amrecover> setdisk /etc
        200 Disk set to /etc.
        No index records for disk for specified date
        If date correct, notify system administrator

        amrecover> history
        200- Dump history for config "DailySet1" host "seagull" disk "/etc"
        201- 2004-08-20 0 DailySet1-002 22
        201- 2004-08-16 0 DailySet1-001 11
        200 Dump history for config "DailySet1" host "seagull" disk "/etc"

        amrecover> setdate 2004-08-20
        200 Working date set to 2004-08-20.
        No index records for cwd on new date
        Setting cwd to mount point


seagull is an HPUX client.  The amindexd debug log shows that the index 
record is being found, but errors are encountered when reading it.  
>From amindexd.20040820142237.debug:

        amindexd: time 14.170: - 2004-08-20 0 DailySet1-002 22
        amindexd: time 14.170: - 2004-08-16 0 DailySet1-001 11
        amindexd: time 14.170: < 200 Disk set to /etc.
        amindexd: time 14.208: > OISD /
        amindexd: time 14.208: uncompress command: /usr/bin/gzip -dc
        '/var/lib/amanda/DailySet1/index/seagull/_etc/20040820_0.gz'
        2>/dev/null | sort >
        '/var/lib/amanda/DailySet1/index/seagull/_etc/20040820_0'
        amindexd: time 14.219: f
        /var/lib/amanda/DailySet1/index/seagull/_etc/20040820_0
        amindexd: time 14.220: < 500 "/" is an invalid directory
        amindexd: time 3418.398: > QUIT


I believe I've tracked this down to a problem with the index database
format the HPUX clients are generating/sending back to the server:

  $ zless /var/lib/amanda/DailySet1/index/seagull/_etc/20040820_0.gz
        10111464545/./
        10111464545/./SnmpAgent.d/
        10111464545/./X11/
        10111464545/./X11/C/
        10111464545/./X11/C.iso88591/
        <...>

According to the amanda docs
(http://www.oops.co.at/AMANDA-docs/indexing.html), the index database
format should be a filename relative to the mount point, starting
with a "/".  Somewhere in the process extraneous (dumpdates(?)) are
being included in the index files generated by HPUX clients.  
For reference, the (working) Linux clients have index databases that
start with at "/".

Also note that if I manually edit
/var/lib/amanda/DailySet1/index/seagull/_etc/20040820_0.gz to remove
the dumpdates(?) so that it becomes
        /
        /SnmpAgent.d/
        /X11/
        /X11/C/
        /X11/C.iso88591/
        <...>
amrecover reads it fine, and is able to recover the data.

Can anyone point me in the right direction to fix the HPUX client
index database format problem?

Thanks,

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Tim Metz, Computer Systems/Network Administrator 
Electrical and Computer Engineering, The University of California, Davis
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=